Alibaba’s Qwen3.8-Max goes open-weight, claiming Claude-rivaling performance

Alibaba has released Qwen3.8-Max, what it calls its largest and most capable model to date, making it broadly available to users. In a blog post Monday, the Chinese tech giant claimed the model’s performance rivals the best systems from US labs like Anthropic and OpenAI, as well as domestic competitors such as Moonshot AI’s Kimi K3. The release follows a preview last month where Alibaba said the model was “second only to Fable 5,” Anthropic’s flagship. Now, the company says its own testing shows Qwen3.8-Max broadly matches—and sometimes exceeds—Fable 5 on benchmarks. On the Arena text model leaderboard, it trails only Fable 5 and three models in Anthropic’s Opus family, and for frontend coding, it’s beaten only by two Claude Opus models and Kimi K3.

The model packs 2.4 trillion parameters, a measure of the settings learned during training. While parameter counts aren’t the sole determinant of performance—Moonshot’s Kimi K3 has 2.8 trillion, and most top US labs keep such figures private—the number signals the scale of the release. More importantly, Alibaba says it will release Qwen3.8-Max’s weights next week. That’s a pivot back to open-weight releases after a brief shift toward proprietary models earlier this year, and it places Alibaba firmly in the camp of Chinese AI firms that have made open weights the norm.

For developers, the open-weight release is the headline. Unlike proprietary APIs from OpenAI or Anthropic, open-weight models give you far more control over the inference stack, enabling fine-tuning, self-hosting, and deeper integration into your own pipelines. The timing is notable: it closely follows Kimi K3’s release, and both ByteDance and MiniMax launched capable video generation models on Friday. The pace of releases from Chinese labs is accelerating, and Qwen3.8-Max appears to be a serious contender, not just in raw benchmarks but in the frontend-coding and visual-analysis categories where it ranks near the top.

The release also lands amid a heated US debate over open-weight models. Reports of a potential crackdown on open tools have been met with industry pushback, with many arguing that access is essential for safety research and competition. Meanwhile, closed-model providers like OpenAI and Anthropic face scrutiny after revealing that their own AI agents were unwittingly involved in cyberattacks, raising questions about the trade-offs of restrictive safety rails. Alibaba’s open-weight move adds fuel to that debate, illustrating the strategic advantage open releases can offer in shaping global AI governance and adoption.